Traditional approaches to tracking user preferences may involve aggregating preference data for many users in a central location and using statistical analysis to identify preference tendencies on a demographic basis. Aggregating potentially sensitive data in a central location, however, may give use to privacy and/or security concerns. Moreover, such an approach may fail to adequately capture individual aspects of user likes and dislikes that may deviate from group demographics.